1 INTRODUCTION

1.1 WHY STUDY ORGANOMETALLIC CHEMISTRY?

Organometallic chemists try to understand how organic molecules or groups interact with compounds of the inorganic elements, chiefl y metals. These elements can be divided into the main group, consisting of the s and p blocks of the periodic table, and the transition elements of the d and f blocks. Main-group organometallics, such as n -BuLi and PhB(OH)2 , have proved so useful for organic synthesis that their leading characteristics are usually extensively covered in organic chemistry courses. Here, we look instead at the transition metals because their chemistry involves the intervention of d and f orbitals that bring into play reaction pathways not readily accessible elsewhere in the periodic table. While main-group organometallics are typically stoichiometric reagents, many of their transition metal analogs are most effective when they act as catalysts. Indeed, the expanding range of applications of catalysis is a COPYRIGHTEDmajor reason for the continued MATERIAL rising interest in organo- metallics. As late as 1975, the majority of organic syntheses had no recourse to transition metals at any stage; in contrast, they now very often appear, almost always as catalysts. Catalysis is also a central prin- ciple of Green Chemistry1 because it helps avoid the waste formation,

The Organometallic Chemistry of the Transition Metals, Sixth Edition. Robert H. Crabtree. © 2014 John Wiley & Sons, Inc. Published 2014 by John Wiley & Sons, Inc. 1 2 INTRODUCTION for example, of Mg salts from Grignard reactions, that tends to accom- pany the use of stoichiometric reagents. The fi eld thus occupies the borderland between organic and inorganic chemistry. The noted organic chemist and Associate Editor of the Journal of Organic Chemistry , Carsten Bolm,2 has published a ringing endorse- ment of organometallic methods as applied to organic synthesis:

In 1989, OMCOS-VI [the 6th International Conference on Organometal- lic Chemistry Directed Toward Organic Synthesis] took place in Florence and . . . left me with the impression that all important transformations could— now or in the future— be performed with the aid of adequately fi ne-tuned metal catalysts. Today, it is safe to say that those early fi ndings were key discoveries for a conceptual revolution that occurred in organic chemistry in recent years. Metal catalysts can be found everywhere, and many synthetic advances are directly linked to . . . developments in cata- lytic chemistry.

Organometallic catalysts have a long industrial history in the produc- tion of organic compounds and polymers. Organometallic chemistry was applied to nickel refi ning as early as the 1880s, when Ludwig Mond showed how crude Ni can be purifi ed with CO to volatilize the Ni in the form of Ni(CO) 4 as a vapor that can subsequently be heated to deposit pure Ni. In a catalytic application dating from the 1930s, Co2 (CO)8 brings about hydroformylation, in which H2 and CO add to an olefi n, such as 1- or 2-butene, to give n-pentanal or n-pentanol, depending on the conditions. A whole series of industrial processes has been developed based on transition metal organometallic catalysts. For example, there is intense activity today in the production of homochiral molecules, in which racemic reagents can be transformed into single pure enantiomers of the product by an asymmetric catalyst. This application is of most sig- nifi cance in the pharmaceutical industry where only one enantiomer of a drug is typically active but the other may even be harmful. Other examples include polymerization of alkenes to give polyethylene and polypropylene, hydrocyanation of butadiene for nylon manufacture, acetic acid manufacture from MeOH and CO, and hydrosilylation to produce silicones and related materials. Beyond the multitude of applications to organic chemistry in indus- try and academia, organometallics are beginning to fi nd applications elsewhere. For example, several of the organic light-emitting diode (OLED) materials recently introduced into cell phone displays rely on organometallic iridium compounds. They are also useful in solid-state light-emitting electrochemical cells (LECs).3 Samsung has a plant that has been producing OLED screens since 2008 that use a cyclometallated COORDINATION CHEMISTRY 3

Ir complex as the red emitter. Cyclometallated Ru complexes may have potential as photosensitizers for solar cells. 4 Organometallic drugs are also on the horizon. Bioinorganic chemistry has traditionally been concerned with classical coordination chemistry—the chemistry of metal ions sur- rounded by N- or O-donor ligands, such as imidazole or acetate— because metalloenzymes typically bind metals via such N or O donors. Recent work has identifi ed a small but growing class of metalloenzymes with organometallic ligands such as CO and CN – in hydrogenases or the remarkable central carbide bound to six Fe atoms in the active site MoFe cluster of nitrogenase. Medicinally useful organometallics, such as the ferrocene-based antimalarial, ferroquine, are also emerging, together with a variety of diagnostic imaging agents. 5 The scientifi c community is increasingly being urged to tackle prob- lems of practical interest.6 In this context, alternative energy research, driven by climate change concerns, 7 and green chemistry, driven by environmental concerns, are rising areas that should also benefi t from organometallic catalysis.8 Solar and wind energy being intermittent, conversion of the resulting electrical energy into a storable fuel is pro- posed. Splitting water into H2 and O2 is the most popular suggestion for converting this electrical energy into chemical energy in the form of H–H bonds, and organometallics are currently being applied as cata- lyst precursors for water splitting.9 Storage of the resulting hydrogen fuel in a convenient form has attracted much attention and will prob- ably require catalysis for the storage and release steps. The recent extreme volatility in rare metal prices has led to “earth-abundant” metals being eagerly sought 10 as replacements for the precious metal catalysts that are most often used today for these and other practically important reactions.

1.2 COORDINATION CHEMISTRY

Even in organometallic compounds, N- or O-donor coligands typical of coordination chemistry are very often present along with C donors. With the rise of such mixed ligand sets, the distinction between coor- dination and organometallic chemistry is becoming blurred, an added reason to look at the principles of coordination chemistry that also underlie the organometallic area. The fundamentals of metal–ligand bonding were fi rst established for coordination compounds by the founder of the fi eld, Alfred Werner (1866–1919). He was able to identify the octahedral geometric preference of CoL 6 complexes without any of the standard spectroscopic or crystallographic techniques.11 4 INTRODUCTION

Central to our modern understanding of both coordination and organometallic compounds are d orbitals. Main-group compounds either have a fi lled d level that is too stable (e.g., Sn) or an empty d level that is too unstable (e.g., C) to participate signifi cantly in bonding. Partial fi lling of the d orbitals imparts the characteristic properties of the transition metals. Some early-transition metal ions with no d elec- trons (e.g., group 4 Ti 4 + ) and some late metals with a fi lled set of 10 (e.g., group 12 Zn 2 + ) more closely resemble main-group elements. Transition metal ions can bind ligands (L) to give a coordination 2 + compound, or complex ML n, as in the familiar aqua ions [M(OH 2 )6 ] (M = V, Cr, Mn, Fe, Co, or Ni). Together with being a subfi eld of organic chemistry, organometallic chemistry can thus also be seen as a subfi eld of coordination chemistry in which the complex contains an M–C bond (e.g., Mo(CO)6 ). In addition to M–C bonds, we include M–L bonds, where L is more electropositive than O, N, and halide (e.g., M–SiR 3 and M–H). These organometallic species tend to be more covalent, and the metal more reduced, than in classical coordination compounds. Typical ligands that usually bind to metal ions in their more reduced, low valent forms are CO, alkenes, and arenes, as in Mo(CO) 6 , Pt(C2 H4 )3 , and (C6 H6 )Cr(CO)3 . Higher valent states are beginning to play a more important role, however, as in hexavalent WMe 6 and pentavalent O = Ir(mesityl)3 (Chapter 15 ).

1.3 WERNER COMPLEXES

3 + In classical Werner complexes, such as [Co(NH3 )6 ] , a relatively high valent metal ion binds to the lone pairs of electronegative donor atoms, typically, O, N, or halide. The M–L bond has a marked polar covalent character, as in L n M–NH3 , where L n represents the other ligands present. The M–NH3 bond consists of the two electrons present in lone pair of free NH3 , but now donated to the metal to form the complex.

Stereochemistry

The most common type of complex, octahedral ML6 , adopts a geometry (1.1 ) based on the Pythagorean octahedron. By occupying the six ver- tices of an octahedron, the ligands can establish appropriate M–L bonding distances, while maximizing their L···L nonbonding distances. For the coordination chemist, it is unfortunate that Pythagoras decided to name his solids after the number of faces rather than the number of vertices. The solid and dashed wedges in 1.1 indicate bonds that point toward or away from us, respectively: WERNER COMPLEXES 5

The assembly of metal and ligands that we call a complex may have a 2− net ionic charge, in which case it is a complex ion (e.g., [PtCl 4 ] ). Together with the counterions, we have a complex salt (e.g., K2 [PtCl4 ]). In some cases, both cation and anion may be complex, as in the pictur- esquely named Magnus ’ green salt [Pt(NH3 )4 ][PtCl4 ], where the square brackets enclose the individual ions. Ligands that have a donor atom with more than one lone pair can often donate one pair to each of two or more metal ions to give poly- nuclear complexes, such as 1.2 (L = PR 3). The bridging group is repre- + sented by the Greek letter μ (mu) as in [Ru2 ( μ -Cl)3 (PR3 )6 ] . Dinuclear 1.2 consists of two octahedra sharing a face containing three chloride bridges.

Chelate Effect Ligands with more than one donor atom, such as ethylenediamine (NH2 CH2 CH2 NH 2, or “en”), can donate both lone pairs to form a chelate ring ( 1.3). The most favorable ring size is fi ve, but six is often seen. Chelating ligands are much less easily displaced from a complex than are comparable monodentate ligands for the reason illustrated in Eq. 1.1 : 6 INTRODUCTION

nn++ [(MNH36 )]+→36 en [()] Men 3 + NH 3 (1.1)

When the reactants release six NH3 molecules in Eq. 1.1 , the total number of particles increases from four to seven. This creates entropy and so favors the chelate. Each chelate ring usually leads to an addi- tional factor of about 10 5 in the equilibrium constant for the reaction. Equilibrium constants for complex formation are usually called forma- tion constants ; the higher the value, the more stable the complex. Chelate ligands can also be polydentate, as in tridentate 1.4 and hexadentate 1.5 . As a tridentate ligand, 1.4 is termed a pincer ligand, a type attracting much recent attention. 12 Ethylenediaminetetracetic acid, (EDTA, 1.5) can take up all six sites of an octahedron and thus completely wrap up many different metal ions. As a common food preservative, EDTA binds free metal ions so that they can no longer catalyze aerial oxidation of the foodstuff. Reactivity in metal complexes usually requires the availability of open sites or at least labile sites at the metal.

Werner ’ s Coordination Theory Alfred Werner developed the modern picture of coordination com- plexes in the 20 years that followed 1893, when, as a young scientist, he proposed that the well-known cobalt ammines (ammonia complexes) have an octahedral structure as in 1.3 and 1.6 .

In doing so, he opposed the standard view that the ligands were bound in chains with the metal at one end (e.g., 1.7 ), as held by everyone else in the fi eld. Naturally, he was opposed by supporters of the stan- dard model, who only went so far as adjusting their model to take WERNER COMPLEXES 7

account of new data. Jørgensen, who led the traditionalists against the Werner insurgency, was not willing to accept that a trivalent metal, Co3 +, could form bonds to more than three groups and so held to the chain theory. At fi rst, as each new “proof” came from Werner, Jør- gensen was able to point to problems or reinterpret the chain theory to fi t the new facts. For example, coordination theory calls for two + isomers of [Co(NH 3 )4 Cl2 ] ( 1.6 and 1.8 ). Up to that time, only a green one had ever been found, now called the trans isomer (1.6 ) because the two Cl ligands occupy opposite vertices of the octahedron. Accord- ing to Werner, a second isomer, 1.8 (cis), then unknown, should have had the Cl ligands in adjacent vertices—he therefore needed to fi nd this isomer. Changing the chloride to nitrite, Werner was indeed able + to obtain both green cis and purple trans isomers of [Co(NH3 )4 (NO2 )2 ] (1.9 and 1.10). Jørgensen quite reasonably—but wrongly—countered this fi nding by saying that the nitrite ligands in the two isomers were simply bound in a different way ( linkage isomers), via N in one case (Co–NO2 ) and O (Co–ONO) in the other (1.11 and 1.12 ). Undis- mayed, Werner then found the green and purple isomers, 1.13 and + 1.14 , of [Co(en) 2 Cl2 ] , in a case where no linkage isomerism was pos- sible. Jørgensen brushed this observation aside by invoking different chain arrangements, as in 1.15 and 1.16 :

8 INTRODUCTION

In 1907, Werner fi nally made the elusive purple isomer of + [Co(NH3 ) 4 Cl 2 ] by an ingenious route (Eq. 1.2 ) via the necessarily cis carbonate [Co(NH3 ) 4(O 2CO)]. Treatment with HCl in the solid state at 0°C liberates CO2 and gives the elusive cis dichloride. Jør- gensen, receiving a sample of this purple complex by mail, fi nally conceded defeat.

(1.2)

+ Werner later resolved optical isomers of the halides [Co(en) 2 X2 ] (1.17 and 1.18), where the isomerism can arise from an octahedral array, but not from a chain. Even this point was challenged on the grounds that only organic compounds could be optically active, and so this activity must come from the organic ligands in some undefi ned way. Werner responded by resolving a complex ( 1.19 ) containing only inorganic ele- ments. This has the extraordinarily high specifi c rotation of 36,000° and required 1000 recrystallizations to resolve.

THE TRANS EFFECT 9

This episode provides general conclusions of importance: some of our current ideas are likely to be wrong—we just do not know which ones. The literature must thus be read critically with an eye for possible fl aws in the results, inferences, or arguments. Nugent has reviewed a series of ideas, once generally held, that subsequently fell from grace. 13 Another lesson from Werner is that we must take objections seriously and devise critical experiments that distinguish between possible theories, not merely ones that confi rm our own ideas.

1.4 THE TRANS EFFECT

We now move from complexes of Co 3 + , or “Co(III),” to the case of Pt(II), where the II and III refer to the +2 or +3 oxidation states (Section 2.4 ) of the metal. Pt(II) is four coordinate and adopts a square planar geometry, as in 1.20 . These complexes can react with incoming ligands, L i , to replace an existing ligand L in a substitution reaction. Where a choice exists between two possible geometries of the product, as in Eq. 1.3 and Eq. 1.4 , the outcome is governed by the trans effect. For example, in the second step of Eq. 1.3 , the NH3 does not replace the Cl trans to NH 3 , but only the Cl trans to Cl. This observation means that Cl is a higher trans effect ligand than NH 3. Once again, in Eq. 1.4 , the NH3 trans to Cl is displaced, not the one trans to NH3 .

Ligands, Lt , that are more effective at labilizing a ligand trans to them- selves have a higher trans effect. We see the reason in Sections 4.3 – 4.4 , but for the moment, only note that the effect is very marked for Pt(II), and that the highest trans effect ligands either (i) form strong σ bonds, such as Lt = H − or Me − , or (ii) are strong π acceptors, such as Lt = CO, C2 H 4 , or (iii) have polarizable period 3 or higher p block elements as donor as in S-bound thiourea, {(NH 2 ) 2 CS or “tu”}. One of the highest − 14 trans effect ligands of all, CF 3 , falls into classes (i) and (ii). High trans effect Lt ligands also lengthen and weaken trans M–L bonds, as shown in X-ray crystallography by an increase in the M–L distance or in nuclear magnetic resonance (NMR) spectroscopy by a decrease in the M,L coupling (Section 10.4 ), or in the IR (infrared) spectrum (Section 10.8 ) by a decrease in the ν(M–L) stretching fre- quency. When L t changes the ground-state thermodynamic properties 10 INTRODUCTION of a complex in one of these ways, we use the term trans infl uence to distinguish the situation from the trans effect proper in which L t accel- erates the rate of substitution, a kinetic effect. An important application of the trans effect is the synthesis of spe- cifi c isomers of coordination compounds. Equation 1.3 and Equation1.4 show how the cis and trans isomers of Pt(NH 3 )2 Cl2 can be prepared selectively by taking advantage of the trans effect order Cl > NH3 , where Lt = Cl. This example is also of practical interest because the cis isomer is a very important antitumor drug (Section 16.5 ), but the trans isomer is toxic.

(1.3)

(1.4)

A trans effect series for a typical Pt(II) system is given below. The order can change somewhat for different metals and oxidation states.

1.5 SOFT VERSUS HARD LIGANDS

Ligands may be hard or soft depending on their propensity for ionic (hard) or covalent (soft) bonding. Likewise, metals can also be hard or soft. The favored, well-matched combinations are a hard ligand with a hard metal and a soft ligand with a soft metal; hard–soft combinations are disfavored because of the mismatch of bonding preferences.15 Table 1.1 shows formation constants for different metal ion–halide ligand combinations, 15 where large positive numbers refl ect strong binding. The hardest halide is F − because it is small, diffi cult to polarize, and forms predominantly ionic bonds. It binds best to a hard cation, H +, also small and diffi cult to polarize. This hard–hard combination therefore leads to strong bonding and HF is a weak acid (pK a + 3). Iodide is the softest halide because it is large, easy to polarize, and forms predominantly covalent bonds. It binds best to a soft cation, Hg 2 + , also large and easy to polarize. In this context, high polarizability means that electrons from each partner readily engage in covalent bonding. THE CRYSTAL FIELD 11

TABLE 1.1 Hard and Soft Acids and Bases: Some Formation Constants a Ligand (Base) Metal Ion (Acid) F − (Hard) Cl − Br − I − (Soft) H + (hard) 3 –7 –9 –9.5 Zn2 + 0.7 –0.2 –0.6 –1.3 Cu 2 + 0.05 0.05 –0.03 – Hg 2 + (soft) 1.03 6.74 8.94 12.87 aThe values are the negative logarithms of the equilibrium constant for [M.aq] n + + X − ⇋ [MX.aq]( n − 1)+ and show how H + and Zn2 + are hard acids, forming stronger complexes with F − than with Cl − , Br −, or I − . Cu 2 + is a borderline case, and Hg 2 + is a very soft acid, forming much stronger complexes with the more polarizable halide ions.

The Hg2 + /I − soft–soft combination is therefore a very good one—by far the best in the table—and dominated by covalent bonding. HI, a mis- matched pairing, produces a strong acid (pK a –9.5). Soft bases either have lone pairs on atoms of the second or later row − − of the periodic table (e.g., Cl , Br , and PPh 3 ) or have double or triple − bonds (e.g., CN , C2 H4 , and benzene) directly adjacent to the donor atom. Soft acids can come from the second or later row of the periodic table (e.g., Hg 2 + ) or contain atoms that are relatively electropositive (e.g., BH3 ) or are metals in a low (≤ 2) oxidation state (e.g., Ni(0), Re(I), Pt(II), and Ti(II) ). Organometallic chemistry is dominated by soft–soft interactions, as in metal carbonyl, alkene, and arene chemistry, while traditional coordination chemistry involves harder metals and ligands.

1.6 THE CRYSTAL FIELD

An important advance in understanding the spectra, structure, and magnetism of transition metal complexes is provided by crystal fi eld theory (CFT) which shows how the d orbitals of the transition metal are affected by the ligands. CFT is based on the very simple model that these ligands act as negative charges, hence crystal fi eld . For Cl − , this is the negative charge on the ion, and for NH3 , it is the N lone pair, a local concentration of negative charge. If the metal ion is isolated in space, then the fi ve d orbitals are degenerate (have the same energy). As the six ligands approach from the octahedral directions ±x , ± y , and ± z , the d orbitals take the form shown in Fig. 1.1 . The d orbitals that point along the axes toward the incoming L groups () dd()xy22− and z 2 are destabilized by the negative charge of the ligands and move to higher energy. Those that point away from L (d xy , d yz , and d xz ) are less destabilized. 12 INTRODUCTION

eg

dz2 dx2 − y2

t2g

n+ n+ M ML6 dxy dyz dxz Octahedral FIGURE 1.1 Effect on the d orbitals of bringing up six ligands along the ± x , ±y , and ±z directions. In this fi gure, shading represents the symmetry (not the occupation) of the d orbitals; shaded parts have the same sign of ψ . For con- venience, energies are shown relative to the average d -orbital energy.

The most strongly destabilized pair of orbitals are labeled e g , from their symmetry, or more simply as d σ , because they point directly along the M–L directions. The set of three more stable orbitals has the label t 2 g , or simply d π —they point between the ligand directions but can still form π bonds with suitable ligands. The energy difference between the d σ and d π set, the crystal fi eld splitting, is labeled Δ (or sometimes 10Dq ) and depends on the value of the effective negative charges and therefore on the nature of the ligands. A higher Δ means we have stronger M–L bonds.

High Spin versus Low Spin In group 9 cobalt, the nine valence electrons have the confi guration [Ar]4s 2 3d 7, but only in the free atom. Once a complex forms, however, the 3 d orbitals become more stable than the 4s as a result of M–L bonding, and the effective electron confi guration becomes [Ar]4s 0 3d 9 for a Co(0) complex, or [Ar]3s 0 4d 6 for Co(III), usually shortened to d 9 and d 6, respectively. The 4 s orbital is now less stable than 3 d because, pointing as it does in all directions, the 4 s suffers CFT repulsion from all the ligands in any Co complex, while the 3 d orbitals only interact with a subset of the ligands in the case of the d σ set or, even less desta- bilizing, point between the ligands in the case of the d π set. THE CRYSTAL FIELD 13

FIGURE 1.2 In a d 6 metal ion, both low- and high-spin complexes are pos- sible depending on the value of Δ. A high Δ leads to the low-spin form ( left ).

This crystal fi eld picture explains why Werner ’ s d 6 Co3 + has such a strong octahedral preference. Its six electrons just fi ll the three low- lying d π orbitals of the octahedral crystal fi eld diagram and leave the higher energy d σ orbitals empty. Stabilizing the electrons in a mole- cule is equivalent to stabilizing the molecule itself. Octahedral d 6 is by far the commonest type of metal complex in all of organometallic chemistry, as in Mo(0), Re(I), Fe(II), Ir(III), and Pt(IV) complexes. In spite of the high tendency to spin-pair the electrons in the d 6 con- 60 fi guration (to give the common low-spin form te 2gg), if the ligand fi eld splitting is small enough, the electrons may rearrange to give the rare 42 high-spin form t 2gge . In high spin (h.s.), all the unpaired spins are aligned (Fig. 1.2 ), as called for in the free ion by Hund’ s rule. Two spin-paired ( ↑ ↓) electrons in the same orbital suffer increased electron–electron repulsion than if they each occupied a separate orbital (↑ )(↑ ). The h.s. form thus benefi ts from having fewer electrons paired up in this way. Unless Δ is very small, however, the energy gained by dropping from the e g to the t2 g level to go from h.s. to l.s. is suffi cient to overcome the e – —e – repulsion from spin pairing, resulting in an l.s. state. The spin state is found from the magnetic moment, determined by comparing the apparent weight of a sample of the complex in the pres- ence and absence of a magnetic fi eld gradient. In l.s. d6 , the complex is diamagnetic and very weakly repelled by the fi eld, as is found for most organic compounds, also spin paired. On the other hand, the h.s. form is paramagnetic , in which case it is attracted into the fi eld because of the magnetic fi eld of the unpaired electrons. The complex does not itself form a permanent magnet as can a piece of iron or nickel—this is ferromagnetism —because the spins are not aligned in the crystal in the absence of an external fi eld, but they do respond to the external fi eld 14 INTRODUCTION by aligning against the applied fi eld when we put them in a magnetic fi eld to measure the magnetic moment. With their high-fi eld ligands, even d n confi gurations and high Δ , the majority of organometallic complexes are diamagnetic, but interest in paramagnetic organometallics (Chapter 15 ) is on the rise. Mononuclear 5 complexes with an uneven number of electrons, such as d V(CO)6 , cannot avoid paramagnetism even in the low-spin case. For even d n confi gurations, high spin is more often seen for the fi rst row metals, where Δ tends to be smaller than in the later rows. Sometimes, the low- and high-spin isomers have almost exactly the same energy. Each state can now be populated in a temperature-dependent ratio, as in Fe(dpe) 2 Cl 2 . Different spin states have different structures and reactiv- ity and, unlike resonance forms, may have a separate existence.

Inert versus Labile Coordination In octahedral d 7 , one electron has to go into the higher energy, less 61 stable e g level to give the low-spin te 2gg confi guration and make the complex paramagnetic (Fig. 1.3 ). The crystal fi eld stabilization energy (CFSE) of such a system is therefore less than for low-spin d6 , where we can put all the electrons into the more stable t 2g level. This is refl ected in the chemistry of octahedral d 7 ions, such as Co(II), that are orders of magnitude more reactive in ligand dissociation than their d 6 analogs because the e g or d σ levels are M–L σ -antibonding (Section 1.7 ). Werner studied Co(III) precisely because the ligands tend to stay put. This is why Co(III) and other low-spin, octahedral d 6 ions are consid- ered coordinatively inert . A half-fi lled t 2g level is also stable, so octahe- dral d 3 is also coordination inert, as seen for Cr(III). On the other hand,

FIGURE 1.3 A d 7 octahedral ion is paramagnetic in both the low-spin ( left ) and high-spin (right ) forms. THE CRYSTAL FIELD 15

Co(II), Cr(II) and all other non- d 6 low-spin and non- d 3 ions are con- sidered coordinatively labile. Second- and third-row transition metals form much more inert complexes than the fi rst-row because of their higher Δ and CFSE.

Jahn–Teller Distortion The lability of some coordination-labile ions, such as d 7 low spin, is aided by a geometrical distortion. This Jahn–Teller (J–T) distortion occurs whenever the individual orbitals in a set of orbitals of the same energy—degenerate orbitals—are unequally occupied. For a pair of degenerate e g orbitals, this requires occupation by one or three elec- 7 trons. Such is the case for low-spin d , where only one of the e g orbitals is half-fi lled (Fig. 1.4 ). In such a case, a pair of ligands that lie along one axis—call this the z axis—either shows an elongation or a contraction

FIGURE 1.4 Jahn–Teller distortions for d 7 low-spin. Uneven occupation of

the d σ orbitals leads to a distortion in which either the xy ML4 ligand set (left) or the z ML2 ligand set (right) shows an M–L elongation because of electron– electron repulsions. Minor splitting also occurs in the d π set. These types of diagrams do not show absolute energies—instead, the “center of gravity” of the orbital pattern is artifi cially kept the same for clarity of exposition. 16 INTRODUCTION of the M–L distances relative to those in the xy plane, depending on whether the () dd()xy22− or z 2 orbital is half-occupied. On crystal fi eld ideas, the electron in the half-fi lled d z2 orbital repels the ligands that lie on the z axis, making these M–L bonds longer; if the d ()xy22− orbital is half occupied, the bonds in the xy plane are longer. This distortion promotes ligand dissociation because two or four of the M–L distances are already elongated and weakened relative to the 6 d low-spin comparison case. A J-T distortion also occurs if the t 2 g set of three orbitals are unevenly occupied with 1, 2, 4, or 5 electrons 6 in t 2g , as in d high spin (Fig. 1.2 , right), but the distortion is now smaller because these t 2g orbitals do not point directly at the ligands. The J-T distortion splits the d orbitals to give a net electron stabi- lization relative to the pure octahedron. This is seen in Fig. 1.4 , where the seventh electron is stabilized whichever of the two distortions, axial or equatorial, is favored.

Low- versus High-Field Ligands

Light absorption at an energy that corresponds to the d π–d σ splitting, Δ , leads to temporary promotion of a d π electron to the d σ level, typi- cally giving d block ions their bright colors. The UV-visible spectrum of the complex can then give a direct measure of Δ and therefore of the crystal fi eld strength of the ligands. High-fi eld ligands, such as CO – and C 2 H4 , lead to a large Δ . Low-fi eld ligands, such as F or H 2 O, can give such a low Δ that even the d 6 confi guration can become high spin and thus paramagnetic (Fig. 1.2 , right side). The spectrochemical series ranks ligands in order of increasing Δ . The range extends from weak-fi eld π -donor ligands, such as halide and H2 O with low Δ , to strong-fi eld π-acceptor ligands, such as CO that give high Δ (Section 1.6 ). These π effects are not the whole story, 16 however, because H, although not a π -bonding ligand, nev- ertheless is a very strong-fi eld ligand from its very strong M–H σ bonds (Section 1.8 ). – – – – – I < Br < Cl < F < H2O < NH3 < PPh3 < CO, H < SnCl3 low high donor acceptor/strong donor Hydrides and carbonyls, with their strong M–L bonds (L = H, CO) and high Δ , are most often diamagnetic. High-fi eld ligands resemble high trans-effect ligands in forming strong σ and/or π bonds, but the precise order differs a little in the two series and for different sets of complexes. THE CRYSTAL FIELD 17

Magnetism and Nuclearity

n 7 A d confi guration where n is odd, such as in d [Re(CO)3 (PCy3 ) 2 ], leads to paramagnetism in a mononuclear complex. In a dinuclear complex, however, the odd electron on each metal can now pair up in forming 7 7 the M–M bond, as in the diamagnetic d –d dimer, [(OC)5 Re–Re(CO)5 ]. Mononuclear complexes with an even d n confi guration can be diamag- netic or paramagnetic depending on whether they are low or high spin. The practical diffi culties of working with paramagnetic complexes, such as the complexity of analyzing their NMR spectra—if indeed any NMR spectrum is detectable at all (Section 10.2 )—has slowed research in the area. Paramagnetism is more common in the fi rst row because their smaller Δ favors high-spin species. The rising cost of the precious metals and the infl uence of green chemistry has made us take much more recent interest in the cheaper fi rst-row metals.

Other Geometries After octahedral, the next most common geometries are three types of 4- or 5-coordination: tetrahedral, square pyramidal and square planar. Tetrahedral is seen for d0 , d 5 (h.s.), and d 10, where we have symmetrical occupation of all the d orbitals, each having zero, one, or two electrons as in Ti(IV), Mn(II), and Pt(0). Since ligand fi eld effects require un sym- metrical d orbital occupation, such effects no longer apply and a tetra- hedral geometry is adopted on purely steric grounds. The orbital pattern—three up, two down (Fig. 1.5 , top)—is the opposite of that for octahedral geometry, and Δ tet is smaller than Δ oct, all else being equal, because we now only have four ligands rather than six to split the d orbitals. Tetrahedral geometry is typical for d4 (low spin), as in Re(III), where only the low-lying pair of d orbitals is occupied. The important square planar geometry, formally derived from an octahedron by removing a pair of trans ligands along the ± z axis, has a more complex splitting pattern (Fig. 1.5 , lower). This derives from the octahedral pattern by pushing the distortion of Fig. 1.4 (right) to the limit. The big splitting, Δ in Fig. 1.5 (right), separates the two highest- energy orbitals. The square planar geometry is most often seen for d 8 (l.s.), as in Pd(II), where only the highest energy orbital remains unoc- cupied. It is also common for paramagnetic d 9 , as in Cu(II). In square pyramidal geometry, only one axial L is removed from octahedral. Holding the geometry and ligand set fi xed, different metal ions can have very different values of Δ . For example, fi rst-row metals and metals in a low oxidation state tend to have low Δ, while second- and third-row metals and metals in a high oxidation state tend to have high 18 INTRODUCTION

FIGURE 1.5 Crystal fi eld splitting patterns for the common four- and fi ve- coordinate geometries: tetrahedral, square pyramidal, and square planar. For the square pyramidal and square planar arrangements, the z axis is convention- ally taken to be perpendicular to the L4 plane. Octahedral geometry is expected for d6 while square planar and square pyramidal are preferred in d 8; the Δ HOMO–LUMO splittings shown apply to those d n confi gurations.

Δ. The trend is illustrated by the spectrochemical series of metal ions in order of increasing Δ :

Second- and particularly third-row metals tend to have a higher Δ than fi rst-row metals thus have stronger M–L bonds, give more thermally stable complexes that are also more likely to be diamagnetic. Higher THE LIGAND FIELD 19

oxidation states of a given metal also tend to produce higher Δ , enhanc- ing these trends, but for a fair comparison, we would need to keep the same M and L n in different oxidation states. This is rarely the case, because low oxidation state metals are usually found with strong-fi eld ligands that tend to give a high Δ (see the spectrochemical series of ligands earlier) and high oxidation state metals are usually most accessible with weak-fi eld ligands that tend to give a low Δ . The oxidation state trend is therefore partially counteracted by the change in ligand preferences.

Isoconfi gurational Ions Ions of the same d n confi guration show important similarities indepen- dent of the identity of the element. This means that d 6 Co(III) is closer in many properties to d6 Fe(II) than to d7 Co(II). The variable valency of the transition metals leads to many cases of isoconfi gurational ions, and this idea helps us predict new complexes from the existence of isoconfi gurational analogs. Numerous analogies of this type have been established for the pair Ir(III) and Ru(II), for example.

1.7 THE LIGAND FIELD

The crystal fi eld picture gives a useful qualitative understanding, but for a more complete picture, we turn to the more sophisticated ligand fi eld theory (LFT), really a conventional molecular orbital, or MO, picture. In this model (Fig. 1.6 ), we consider the s , the three p , and the fi ve d orbitals of the valence shell of the isolated ion, as well as the six lone-pair orbitals of a set of pure σ -donor ligands in an octahedron around the metal. Six of the metal orbitals, the s , the three p , and the two d σ , the dsp σ set, fi nd symmetry matches in the six ligand lone-pair orbitals. In combining the six metal orbitals with the six ligand orbitals, we make a bonding set of six (the M–L σ bonds) that are stabilized, and an antibonding set of six (the M–L σ * levels) that are destabilized. The remaining three d orbitals, the d π set, do not overlap with the ligand orbitals and remain nonbonding, somewhat resembling lone pairs in p block compounds. In a d 6 ion, we have 6e from Co3 + and 12e from the six :NH3 ligands, giving 18e in all. This means that all the levels up to and including the d π set are fi lled, and the M–L σ * levels remain unfi lled—the most favorable situation for high stability. Note that we can identify the familiar crystal fi eld d orbital splitting pattern in the d π set and two of the M–L σ * levels. The Δ splitting increases as the strength of the M–L σ bonds increases, so bond strength is analogous to the effective charge in the crystal fi eld model. In the ligand fi eld 20 INTRODUCTION

FIGURE 1.6 Molecular orbital, or ligand fi eld picture, of M–L bonding in an octahedral ML6 complex. The box contains the d orbitals that are fi lled with n electrons to give the d n electron confi guration. The star denotes antibonding. picture, one class of high-fi eld ligands form strong σ bonds, for example, H or CH 3 . We can now see that the d σ orbital of the crystal fi eld picture becomes an M–L σ -antibonding orbital in the ligand fi eld model. The L lone pairs in the free ligand become bonding pairs shared between L and M when the M–L σ bonds are formed; these are the six lowest orbitals in Fig. 1.6 and are always completely fi lled with 12e . Each M–L σ -bonding MO is formed by the combination of the ligand lone pair, L( σ), with M( d σ ), and has both M and L character, but L( σ ) predominates. Any MO more closely resembles the parent atomic orbital that lies closest to it in energy, and L(σ ) almost always lies below M(d σ ) and therefore closer to the M–L σ -bonding orbitals. Electrons that were purely L lone pairs in free L now gain some metal character in ML 6 ; in other words, the L( σ ) lone pairs are partially transferred to the metal. As L becomes more basic, the energy of the L( σ ) orbital increases together with the extent of lone pair transfer. An orbital that moves to higher energy moves higher in the MO diagram and tends to THE sdn MODEL AND HYPERVALENCY 21

occupy a larger volume of space; any electrons it contains become less stable and more available for chemical bonding or removal by electron loss in any oxidation. Ligands are generally nucleophilic because they have high-lying lone pair electrons available, while a metal ion is electrophilic because it has low-lying empty d orbitals available. A nucleophilic ligand, a lone-pair donor, can thus attack an electrophilic metal, a lone pair acceptor, to give a metal complex. Metal ions can accept multiple lone pairs so that the complex formed is ML n (n = 2–9).

1.8 THE sdn MODEL AND HYPERVALENCY

The ligand fi eld model is currently being challenged by the sdn model. 17 This considers the np orbital as being ineffective in M–L bonding owing to poor overlap and mismatched energies and proposes that only the ns and fi ve (n – 1)d orbitals contribute, n being 4, 5, and 6 for the fi rst-, second-, and third-row d block metals. For example, photoelectron 3 spectroscopy shows that Me2 TiCl2 has sd hybridization, not the famil- 3 18 6 iar sp hybridization as in Me2 CCl2 . If so, one might expect d metal complexes to prefer a 12-valence electron count, not 18e, since 12e would entirely fi ll the sd 5 set. This would, however, wrongly lead us to expect Mo(CO)3 rather than the observed Mo(CO)6 . To account for the additional bonding power of Mo(CO)3 , hypervalency is invoked. Hypervalency, the ability of an element to exceed the valence elec- tron count normally appropriate for the orbitals that are available, is best established in the main-group elements, such as sulfur, where an octet of eight valence electrons is appropriate for its single s and three p orbitals. In hypervalent SF6 , for example, six electrons come from S and one each from the six F atoms for a total of 12 valence electrons, greatly exceeding the expected octet. The modern theory of hyperva- lency avoids the earlier idea, now exploded, that empty d orbitals (3 d orbitals for S) allow the atom to house the excess electrons. Hypervalent bonding is most simply illustrated for [FHF] − anion, where H has four valence electrons, exceeding its normal maximum of 2e. In [FHF] − , the zero electron H + receives 2e from each of the lone − pairs of the two F anions coordinated to it, thus resembling an ML2 complex. The bonding pattern, shown in Fig. 1.7 , allows the 4e from the two F − to occupy two lower-lying orbitals each having predominant F character—one bonding, one nonbonding—while leaving the highest energy orbital empty. In effect, one 2e bond is spread over two H–F bonds, and the remaining 2e in the nonbonding orbital are predomi- nantly located on F. The resulting 4 electron–3 center (4e–3c) bonding 22 INTRODUCTION

FIGURE 1.7 The four electron–three center (4e–3c) hypervalent bonding model for [FHF] − anion in which the fl uoride ions are considered ligands for the central H +. The bonding and nonbonding orbitals are occupied and the antibonding orbital left vacant. leads to half-order bonding between H and each F, resulting in some- what longer bonds (1.15 Å) than in the corresponding nonhypervalent species, HF (0.92 Å). [FHF] − anion, normally considered as a strong hydrogen-bonded adduct of HF and F − , is here seen as hypervalent. Moving to the heavier p block elements, hypervalent octahedral SF6 , for example, can be considered as having three trans F–S–F units, each bonded via 4e–3c bonds. Main-group hypervalency requires an electronegative ligand, often F or O, that can stabilize the bonding and nonbonding orbitals of Fig. 1.7 . This results in the accumulation of negative charge on the terminal F atoms that are best able to accommodate it. In coordination com- plexes, the ligands are indeed almost always more electronegative than the metal even when we expand the ligand choice beyond F and O to N, P and C donors. To return to Mo(CO)6 , the bonding is explained in terms of three pairs of trans L–M–L hypervalent 4e–3c bonds, formed from sd5 hybrids. This leaves three d orbitals that are set aside for back bonding to CO as the d π set, as in ligand fi eld theory. BACK BONDING 23

Bent ’ s rule, which helps assign geometries for main-group com- pounds, relies on sp 3 hybridization and therefore has to be modifi ed for application to the d block. For example, in Me 2 CCl2 , the Cl–C–Cl angle (108.3°) is less than the C–C–C angle (113.0°), since the more electro- negative Cl atoms elicit a higher contribution from the less stable orbital, in this case, the carbon p orbital. The C–Cl bonds having high p orbital character also have a smaller bond angle, since p orbitals are 90° apart. In Me2 TiCl2 , in contrast, the Cl–Ti–Cl angle (116.7°) is larger than the C–Ti–C angle (106.2°) because the hybridization is now sd 3 and the d orbitals are the more stable members of the sd 3 set. The less electronegative Me substituents now elicit greater Ti d character and have the smaller bond angle.19 The fate of this model depends on whether it fi nds favor in the sci- entifi c community, and we will not use it extensively in what follows. Textbooks can give the impression that everything is settled and agreed upon, but that agreement is only achieved after much argument, leading to an evolution of the community ’ s understanding. Ideas that come to dominate often start out as a minority view. The sd n model may there- fore either fade, fl ower, or be modifi ed in future.

1.9 BACK BONDING

Ligands such as NH3 are good σ donors but insignifi cant π acceptors. CO, in contrast, is a good π acceptor and relatively poor σ donor. Such π -acid ligands are of very great importance in organometallic chemistry. They tend to be very high fi eld ligands and form strong M–L bonds. All have empty orbitals of the right symmetry to overlap with a fi lled d π orbital of the metal: for CO, this acceptor is the empty CO π *. Figure 1.8 shows how overlap takes place to form the M–C π bond. It may

FIGURE 1.8 Overlap between a fi lled metal d π orbital and an empty CO π * orbital to give the π component of the M–CO bond. The shading refers to symmetry of the orbitals. The M–CO σ bond is formed by the donation of a

lone pair on C into an empty d σ orbital on the metal (not shown). 24 INTRODUCTION

FIGURE 1.9 Effect of “turning on” the π interaction between a π -acceptor ligand and the metal. The unoccupied, and relatively unstable, π * orbitals of the ligand are shown on the right. Their effect is to stabilize the fi lled d π orbitals of the complex and so increase Δ . In W(CO)6 , the lowest three orbitals are fi lled. seem paradoxical that an antibonding orbital such as the π *(CO) can form a bond, but this orbital is only antibonding with respect to C and O and can still be bonding with respect to M and C. A second CO π *, oriented out of the image plane, can accept back bonding from a second d π orbital that is similarly oriented. The ligand fi eld diagram of Fig. 1.6 has to be modifi ed when the ligands are π acceptors, such as CO, because we now need to include the CO π * levels (Fig. 1.9 ). The M d π set now interact strongly with the empty CO π * levels to form M-C π bonds. For d 6 complexes, such as W(CO)6 , where the Md π set is fi lled, the Md π electrons now spend some of their time on the ligands by back bonding. Back bonding can occur for a wide variety of M–L bonds as long as L contains a suitable empty orbital. In one type, where the donor atom participates in one or more multiple bonds, the empty orbital is a ligand π*, as is the case for CO or C 2 H4 . As we see in detail in Sections 3.4 and 4.2 , other types of ligand have suitable empty σ * orbitals, as is the case for PF 3 or H 2. On the metal side, back bonding can only happen in d1 or higher confi gurations; a d0 ion such as Ti 4 + cannot back bond and very seldom forms stable complexes with strong π acceptor ligands, such as CO. BACK BONDING 25

Being antibonding, the CO π * levels are high in energy, but they are able to stabilize the d π set by back bonding as shown in Fig. 1.9 . This has two important consequences: (1) The ligand fi eld splitting param- eter Δ rises, explaining why π-bonding ligands have such a strong ligand fi eld and make such strong M–L bonds; and (2) back bonding allows electron density on a low oxidation state metal to make its way back to the π -acid ligands. This applies when low-valent or zero-valent metals form CO complexes. Such metals have a high electron density in the free state and are thus reluctant to accept further electrons from pure σ donors; this is why W(NH3 ) 6 is not known. By back bonding, the metal can get rid of some of this excess electron density and delocalize it over the π-acid ligands. In W(CO) 6, back bonding is so effective that the compound is air stable and relatively unreactive; the CO groups have so stabilized the metal electrons that they have no tendency to be abstracted by an oxidant such as air. In W(PMe 3 )6 , in contrast, back bonding is weak and the complex is reactive and air-unstable. Their structures show that π back donation is a big contributor to the M= C bond in metal carbonyls, making the M= C bond much shorter than an M–C single bond. For example, in CpMo(CO) 3 Me, M–CH3 is 2.38 Å but M = CO is 1.99 Å. A true M–CO single bond would be shorter than 2.38 Å by about 0.07 Å, to allow for the higher s character 3 of sp CO versus sp CH3 , leaving a substantial shortening of 0.32 Å that can be ascribed to back bonding. IR spectroscopy identifi es the CO π * orbital as the acceptor in back bonding. A CO bound only by its carbon lone pair—nonbonding with respect to CO—would have a ν (CO) frequency close to that in free CO. BH3 , a predominant σ acceptor, shows a slight shift of ν (CO) to higher − l − 1 energy in H 3 B-CO: free CO, 2143 cm ; H 3 B–CO, 2178 cm so the shift is + 35 cm − 1 . Metal carbonyls, in contrast, show ν (CO) coordination shifts of hundreds of wavenumbers to lower energy, consistent with the weakening of the C–O bond as the CO π * is partially fi lled by back − 1 − 1 donation; for Cr(CO)6 , ν (CO) is 2000 cm , so the shift is -143 cm . Not only is there a coordination shift, but the shift is larger in cases where we would expect stronger back donation (Table 2.10 ) and ν (CO) is considered a good indicator of metal basicity. In Section 4.2 , we see how the ν (CO) of LNi(CO)3 helps us rank different ligands L in terms of their comparative donor power to M; good donor L ligands make the Ni back donate more strongly into the CO groups. Formation of the M–CO bond weakens the C ≡O bond of free CO. This can still lead to a stable complex as long as the energy gained from the new M–C bond exceeds the loss in C ≡O. Bond weakening in L on binding to M is very common in M–L complexes where back bonding is signifi cant. 26 INTRODUCTION

− + Series of compounds such as [V(CO)6 ] , Cr(CO)6 , and [Mn(CO)6 ] are isoelectronic because, V(–I), Cr(0), and Mn(I) all being d 6 , they have the same number of electrons similarly distributed. Isoelectronic ligands include CO and NO + and CN − , for example. CO and CS are not strictly isoelectronic, but as the difference between O and S only lies in the number of core levels, while the valence shell is the same, the term is often extended to such pairs. A comparison of isoelectronic complexes or ligands can be very useful in looking for similarities and differences. 20

Frontier Orbitals A similar picture holds for a whole series of soft, π acceptor ligands, such as alkenes, alkynes, arenes, carbenes, carbynes, NO, N2 , and PF3 . Each has a fi lled orbital that acts as a σ donor and an empty orbital that acts as a π acceptor. These orbitals are almost always the highest occupied (HOMO ) and lowest unoccupied molecular orbitals (LUMO ) of L, respectively. The HOMO of L is normally a donor to the d σ LUMO of the metal. The ligand LUMO thus accepts back donation from the metal HOMO, a fi lled metal d π orbital. The HOMO and LUMO of each fragment, the so-called fron- tier orbitals, often dominate the bonding between the fragments. Strong interactions between orbitals require not only good overlap but also that the energy separation between them be small. The HOMO of each frag- ment, M and L, is usually closer in energy to the LUMO of the partner fragment than to any other vacant orbital of the partner. Strong bonding is thus expected if the HOMO–LUMO gap of both partners is small. Indeed a small HOMO–LUMO gap for any molecule gives rise to high reactivity. A small HOMO–LUMO gap also makes a ligand soft because it becomes a good π acceptor, and for d 6 , makes the metal soft because it becomes a good π donor.

π -Donor Ligands Ligands such as OR − and F − are π donors as a result of the lone pairs that are left after one lone pair has formed the M–L σ bond. Instead 6 of stabilizing the d π electrons of an octahedral d ion as does a π accep- tor, these d π electrons are now destabilized by what is effectively a repulsion between two fi lled orbitals. This lowers Δ, as shown in Fig. 1.10 , and leads to a weaker M–L bond than in the π -acceptor case, as 6 3- in high-spin d [CoF6 ] . Lone pairs on electronegative atoms such as − − F and RO are much more stable than the M(d π ) level, and this is why they are lower in Fig. 1.10 than are the π * orbitals in Fig. 1.9 . Having more diffuse lone pairs, larger donor atoms pose fewer problems, and − − 6 Cl and R 2 P are much better tolerated by d metals. ELECTRONEUTRALITY 27

FIGURE 1.10 Effect of “turning on” the π interaction between a π -donor ligand and the metal. The occupied, and relatively stable, lone-pair (π ) orbitals

of the ligand are shown on the right. Their effect is to destabilize the fi lled d π orbitals of the complex and so decrease Δ. In d 6, this is effectively a repulsion between two lone pairs, one on the metal and the other on the ligand, and thus unfavorable for M–L bonding. In d 0, this repulsion is no longer present, and the stabilization of the π lone pairs of L becomes a favorable factor for M–L bonding.

0 In sharp contrast, if the metal has empty d π orbitals, as in the d ion 4 + Ti , π donation from the π -donor ligand to the metal d π orbitals now leads to stronger metal–ligand bonding; d 0 metals therefore form par- 2 − ticularly strong bonds with such ligands, as in W(OMe)6 or [TiF6 ] , both also examples of favorable hard metal–hard ligand combinations.

1.10 ELECTRONEUTRALITY

Linus Pauling (1901–1994), a giant of twentieth-century chemistry, pro- posed the electroneutrality principle in which electrons distribute them- selves in polar covalent molecules so that each atomic charge is nearly neutral. In practice, these charges fall in a range from about + 1 to − 1. The nonmetals tend to be negatively charged with N, O, or F being closer to − 1 and Na or Al, being closer to + 1. This implies that elements that have complementary preferred charges should bond best so each can satisfy the other, as in LiF or TiO 2; in contrast, elements with intermediate electronegativity prefer each other, as in H 2 , HgS, and Au–Ag alloy. An isolated Co 3 + ion is far from electroneutral so it prefers good electron 28 INTRODUCTION

2 − donors as ligands, such as O in Co 2 O3 , or NH 3 in the Werner com- plexes. On the other hand, an isolated W(0) atom is already neutral and is thus too electron rich for its electronegativity, so it prefers net electron-attracting ligands, such as CO that can accept electron density by π back donation so that the metal can attain a positive charge.

Oxidation State Trends The d orbitals of transition metals are only fully available for back donation in low oxidation states. Although d 6 Co(III), for example, does have a fi lled d π level, it is unavailable for back bonding—Co(III) there- fore cannot bind CO. The high positive charge of Co(III) contracts all the orbitals with the result that the d π orbital is low in energy and therefore only weakly basic. Likewise, repulsive effects of π donors such as F − and RO − are mild.

Periodic Trends The orbital energies fall as we go from left to right in the transition series. For each step to the right, a proton is added to the nucleus, thus providing an extra positive charge that stabilizes all the orbitals. The earlier metals are more electropositive because it is easier to remove electrons from their less-stable orbitals. The sensitivity of the orbitals to this change is d ∼ s > p because the s orbital, having a maximum electron density at the nucleus, is more stabilized by the added protons than are the p orbitals, with a planar node at the nucleus. The d orbitals are also stabilized because of their lower principal quantum number, as is the case for 3 d versus 4s and 4 p in the valence shell of Fe. The special property of the transition metals is that all three types of orbital are in the valence shell with similar energies so all contribute signifi - cantly to the bonding, only omitting the 4p if the sd n model is adopted. Metal carbonyls, for example, are most stable for groups 4–8 because CO requires back bonding to bind strongly and in the later groups, the needed d π orbitals become too stable to be effective. Organometallic compounds of the electropositive early metals have a higher polar covalent character than in the later metals and thus tend to be more air-sensitive, because they are more subject both to oxidation by O 2 and hydrolysis by H 2 O. There is a sharp difference between d 0 and d2 as in Ti(IV) versus Ti(II): d 0 Ti(IV) cannot back bond at all, while d 2 Ti(II) is a very strong back-bonder because early in the transition series, where d 2 states are most common, the d orbitals are relatively unstable for the reasons 0 mentioned earlier. The d Ti(IV) species (C 5 H5 )2 TiCl 2 therefore does TYPES OF LIGAND 29

not react with CO at all, while the corresponding d 2 Ti(II) fragment, (C5 H5 )2 Ti, forms a very stable monocarbonyl, (C5 H5 )2 Ti(CO), with a low ν (CO) IR frequency, indicating very strong back bonding. Finally, as we go down a given group in the d block from the fi rst to the second row, the outer valence electrons become more shielded from the nucleus by the extra shell of electrons added. They are therefore more easily lost, making the heavier d block element more basic and more capable of attaining high oxidation states. This trend also extends to the third row, but as the f electrons that were added to build up the lanthanide elements are not as effective as s, p, or even d electrons in shielding the valence electrons from the nucleus, there is a smaller change on going from the second to the third row than from the fi rst to the second. Compare, for example, the power- fully oxidizing Cr(VI) in Na 2CrO 4 and Mn(VII) in KMnO 4, with their stable second- and third-row analogs, Na2 MoO 4, Na2 WO4 , and KReO 4; the very weakly oxidizing character of the latter indicates an increased stability for the higher oxidation state. For the same reason, the increase in covalent radii is larger on going from the fi rst to the second row than from the second to the third. This anomaly in atomic radius for the third row is termed the lanthanide contraction . Mononuclear ionic complexes with excessively high positive or nega- tive net ionic charges are not normally seen. The majority of isolable compounds are neutral; net charges of ± 1 are not uncommon, but higher net ionic charges are rare.

1.11 TYPES OF LIGAND

Most ligands are Lewis bases and thus typically neutral or anionic, rarely cationic. Anionic ligands, often represented as X, form polar covalent M–X bonds. In addition to the σ bond, there can also be a π interaction which may be favorable or unfavorable as discussed in Section 1.9 . Among neutral ligands, often denoted L, we fi nd lone-pair donors, such as :CO or :NH 3 , π donors such as C2 H4 , and σ donors such as H2 . The fi rst group—the only type known to Werner—bind via a lone pair. In contrast, π donors bind via donation of a ligand π -bonding electron pair, and σ donors bind via donation of a ligand σ -bonding electron pair to the metal. The relatively weakly basic σ- and π-bonding electrons of σ and π donors would form only very weak M–L bonds if acting alone. Both σ and π donors therefore normally require some back bonding to produce a stable M–L bond. Even so, the strength of binding tends to 30 INTRODUCTION

(b) (a) 2 4 H C 1 3 M M 4 2 C H

FIGURE 1.11 (a) Bonding of a π -bond donor, ethylene, to a metal. Arrow

1 represents electron donation from the fi lled C= C π bond to the empty d σ orbital on the metal; arrow 2 represents the back donation from the fi lled

M(d π ) orbital to the empty C= C π *. (b) Bonding of a σ -bond donor, hydrogen, to a metal. Arrow 3 represents electron donation from the fi lled H–H σ bond to the empty d σ orbital on the metal, and arrow 4 represents the back donation from the fi lled M(d π) orbital to the empty H–H σ *. Only one of the four lobes of the d σ orbital is shown. decrease as we move from lone pair to π bond to σ bond donors, other factors being equal. For the π donor, ethylene, Fig. 1.11 a illustrates how L to M donation from the C= C π orbital to M d σ (arrow 1) is accompanied by back donation from M d π into the C = C π * orbital (arrow 2). For the σ donor, H2 , Fig. 1.11 b shows how L to M donation from the H–H σ orbital to M d σ (arrow 3) is accompanied by back donation from M d π into the H–H σ * orbital (arrow 4). As always, back bonding requires a d 2 or higher electron confi guration and relatively basic M d π electrons, usually found in low oxidation states. Side-on binding of σ and π donors results in short bonding distances to two adjacent ligand atoms. This type of binding is represented as 2 2 η -C2 H4 or η -H2 , where the letter η (often pronounced eeta) denotes the ligand hapticity, the number of adjacent ligand atoms directly bound 21 to the metal. For σ donors such as H2 , forming the M–L σ bond par- tially depletes the H–H σ bond because electrons that were fully engaged in keeping the two H atoms together in free H 2 are now also delocalized over the metal, hence the name two-electron, three-center (2e,3c) bond for this interaction. Back bonding into the H–H σ * causes additional weakening or even breaking of the H–H σ bond because the σ* is antibonding with respect to H–H. Free H2 has an H–H distance of 0.74 Å, but the H–H distances in H2 complexes go all the way from 0.82 to 1.5 Å. Eventually, the H–H bond breaks and a dihydride is formed (Eq. 1.5 ). This is the oxidative addition reaction (see Chapter 6 ). Formation of a σ complex can be thought of as an incomplete oxidative addition, where only the addition part has occurred. Table 1.2 classifi es common ligands by the nature of the M–L σ and π bonds. Both σ and TYPES OF LIGAND 31

TABLE 1.2 Types of Ligand a Strong π Weak π Ligand Acceptor Bonding Strong π Donor b − c − − − − − Lone-pair donor CO, PF3 , CR 2 H , PPh3 , Me , Cl F , OR , NR 2 π -Bonding electron C2 F 4 , O2 C2 H4 , – pair donor RCH = O d σ -Bonding electron Oxidative R3 Si–H, H–H, – e pair donor Addition R3 C–H g f σ - and π -acceptor BF 3 BH3 , CO2 – aLigands are listed in approximate order of π -donor/acceptor power, with acceptors mentioned fi rst. b Fischer carbene (Chapter 11 ). c Ligands like this are considered here as anions rather than radicals. d Can also bind via an oxygen lone pair (Eq. 1.6 ). e Oxidative addition occurs when σ -bond donors bind very strongly (Eq. 1.5 ). g R a r e . f When bound η1 via C.

π bonds bind side-on to metals when they act as ligands. Alkane C–H bonds behave similarly.22 H H LnM + H2 LnM LnM H H (1.5) complex oxidative addition product

Lewis acids such as BF3 can be ligands by accepting a basic electron pair from the metal (Ln M: → BF3 ), in which case the ligand contributes nothing to the metal electron count: BF3 is also a strong π -acceptor for back bonding from M d π orbitals via the σ* orbitals, as discussed for PF3 in Section 4.2 .

Ambidentate Ligands Alternate types of electron pair are sometimes available for bonding. For example, aldehydes have both a C= O π bond and oxygen lone pairs. As π -bond donors, aldehydes bind side-on (Eq. 1.6 , 1.21a ) like ethylene, but as lone-pair donors, they can alternatively bind end-on (1.21b ). Thiocyanate, SCN -, can bind via N in a linear fashion (Eq. 1.7 , 1.22a ), or via S, in which case the ligand is bent ( 1.22b); in some cases, both forms are isolable. 23 32 INTRODUCTION

(1.6)

(1.7)

II 2 + The {(NH3 ) 5Os } fragment in Eq. 1.8 is a very strong π donor because Os(II) is soft and NH 3 is not a π-acceptor; the π-basic Os thus prefers to bind to the π acceptor aromatic C= C bond of aniline, not to the nitrogen. Oxidation to OsIII causes a sharp falloff in π - donor power because the extra positive charge stabilizes the d orbit- als, and the Os(III) complex slowly rearranges to the N-bound aniline form.24 This illustrates how the electronic character of a metal can be altered by changing the ligand set and oxidation state; soft Os(II) binds to the soft C = C bond and hard Os(III) binds to the hard ArNH2 group.

(1.8)

Figure 1.12 shows the typical ligands found for different oxidation states of Re, an element with a very wide range of accessible states. Low OS complexes are stabilized by multiple π -acceptor CO ligands, intermedi- ate OSs by less π-acceptor phosphines, high OS by σ -donor anionic ligands such as Me, and very high OS by O or F, ligands that are both σ donor and π donor. The dipyridyl phosphine ligand of Eq. 1.9 shows two distinct binding modes, depending on the conditions and anion present.25

(1.9) TYPES OF LIGAND 33

FIGURE 1.12 Some Re complexes showing typical variation of ligand type with oxidation state (OS): hard ligands with high OS and soft ligands with low OS.

Actor and Spectator Ligands Actor ligands associate, dissociate or react in some way. They are particularly important in catalytic reactions, when they bind to the metal and engage in reactions that lead to release of a product mol- ecule. In hydrogenation, for example, H 2 and ethylene can associate to give [Ln MH2 (C2 H4 )] intermediates that go through a cycle of reac- tions (Section 9.3 ) that leads to release of the hydrogenation product, C2 H6 . Spectator ligands remain unchanged during chemical transforma- tions but still play an important role by tuning the properties of the metal to enhance desired characteristics. For example, in the extensive + chemistry of [CpFe(CO)2 X] and [CpFe(CO)2 L] (Cp = cyclopentadi- enyl; X = anion; L = neutral ligand), the {CpFe(CO)2 } fragment remains intact. The spectators impart solubility, stabilize Fe(II), and infl uence the electronic and steric properties of the complex. It is an art to pick suitable spectator ligand sets to elicit desired properties. Apparently small changes in ligand can entirely change the chemistry. For example, PPh3 is an exceptionally useful ligand, while the appar- ently similar NPh3 , BiPh3 , and P(C6 F5 )3 are of very little use. The hard N-donor, NPh 3, is very different from PPh3 ; the Bi-Ph bond is too easily cleaved for BiPh3 to be a reliable spectator; and the electron- withdrawing C6 F5 substituents of P(C 6 F5 )3 completely deactivate the P lone pair. Steric size sets the maximum number of ligands, n , that can fi t around a given metal in a d block MLn complex. Typical n values depend on the size of the ligand: H, 9; CO, 7; PMe 3, 6; PPh 3, 4; P(C 6 H11 )3 , 2, and only in a trans arrangement; a few ligands are so big that n = 1, for 34 INTRODUCTION

FIGURE 1.13 A selection of common ligands with different binding preferences. example, X-Phos (4.11 ). If a big spectator ligand can occupy no more than n sites when the metal has m sites available, then m – n sites are kept open for smaller actor ligands. Multidentate spectator ligands can have the n donor atoms arranged in specifi c patterns and geometries, making the m – n available sites take up a complementary geometry. A small sample of such ligands is shown in Fig. 1.13 . The tridentate ligands can bind to an octahedron either in a mer (meridonal) fashion 1.23 or fac (facial) 1.24 , or in some cases, in both ways. Ligands that normally bind in terdentate mer fashion are pincers. Not only do these benefi t from the chelate effect, but they also allow us to control the binding at three sites of an octahedron, leaving three mer sites acces- sible to reagents. Tetradentate ligands, such as 1.25 can also prove useful, in this case by stabilizing the unusual Pd(III) oxidation state.26 The choice of ligand is an art because subtle stereoelectronic effects, still not fully under- stood, can play an important role. Ligands 1.26 and 1.27 (Fig. 1.13 ) impart substantially different properties to their complexes in spite of their apparent similarity, probably as a result of the greater fl exibility of the three-carbon linker in 1.27 . TYPES OF LIGAND 35

Actor ligands may allow isolation of a stable material as a precursor to a reactive species only formed after departure of the actor, that species either being too reactive to isolate or not otherwise easily accessible. A classic example is chelating 1,5-cyclooctadiene (cod) that binds to + Rh(I) or Ir(I) in the [(cod)M(PR 3 )2 ] hydrogenation catalysts (1.28 ). Under H 2, the cod is hydrogenated to free cyclooctane, liberating + {M(PR3 )2 } as the active catalyst. Cp* is a reliable spectator, except under strongly oxidative conditions, when it can degrade and become an actor. For example, the Cp* in Cp*Ir(dipy)Cl is oxidatively removed − with Ce(IV) or IO 4 to give a homogeneous coordination catalyst capable of oxidizing water or C–H bonds. 27 Similarly, the Cp* in [Cp*Ir(OH2 )3 ]SO4 is oxidatively degraded under electrochemical oxi- dation to yield a heterogeneous water oxidation catalyst that deposits on the electrode.28

Multifunctional Ligands29 These more sophisticated ligands are increasingly being seen. Beyond the simple metal-binding function, numerous additional functionalities can also be incorporated. Some ligands reversibly bind protons, altering their donor properties; others have hydrogen bonding functionality for molecular recognition. Sometimes, a complex can be oxidized or reduced, but the resulting radical is ligand centered so that the metal oxidation state is unchanged.

Organometallic versus Coordination Compounds Originally, the presence of any M–C bonds made a metal complex organometallic—their absence made it a coordination compound. Electronegativity differences (Δ EN) between M and the donor atom 36 INTRODUCTION in L were invoked. Organometallic M–L bonds, such as M–CH3 , typically have a lower Δ EN and are thus more covalent than bonds with greater Δ EN and more ionic character, such as the M–N or M–O bonds typical of coordination complexes. Mixed ligand sets are now much more common, making sharp distinctions less helpful. Ligands such as H, SiR 3, or PR 3 are now regarded as organometallic because Δ EN is low and covalency predominates. In the key subfi eld of cataly- sis, coordination compounds have proved as useful as organometal- lics. For Wilkinson’ s catalyst, [RhCl(PPh 3) 3], one of the most important compounds in the history of the fi eld (see Chapter 9 ), M–C bonds are only present in the intermediates formed during the catalytic cycle. Likewise, in CH activation (Section 12.4 ), many of the catalysts involved are again coordination compounds that operate via organometallic intermediates (e.g., [ReH7 (PPh 3) 2 ] or K2 [PtCl 4 ]). In an increasing number of cases, such as the metal oxo mechanism for CH activation (Sections 12.4 and 14.7 ), no M–C bonds are ever present, even in reaction intermediates. Today, the organometallic/ coordination distinction is therefore losing importance. While still emphasizing traditional organometallics, we therefore do not hesi- tate to cross into coordination chemistry territory on occasion, par- ticularly in Chapters 14 – 16 .

• High trans effect ligands such as H or CO labilize ligands that are trans to themselves. • In CFT (Section 1.6 ), the d -orbital splitting, Δ, and e − occupation determine the properties of the complex.

• Hard ligands, such as NH 3, have fi rst-row donor atoms and no multiple bonds; soft ligands, such as PR 3 or CO, have second-row donors or multiple bonds. • Ligands donate electrons from their HOMO and accept them into their LUMO (p. 26). LFT (Section 1.7 ) identifi es the d σ orbitals as M–L antibonding. • M–L π bonding strongly affects Δ and thus the strength of M–L bonding (Fig. 1.8 , Fig. 1.9 , and Fig. 1.10 ). • Ligands can bind via lone pairs, π bonding e − pairs or σ bonding e − pairs (Table 1.2 ). • Octahedral d 3 and d 6 are coordination inert and slow to dissociate a ligand. REFERENCES 37

REFERENCES

1. R. A. Sheldon , Chem. Comm. , 2008 , 3352 ; P. Anastas , N. Eghbali , Chem. Soc. Rev. , 39 , 301 , 2010 . 2. C. Bolm , Org. Lett. , 14 , 2926 , 2012 . 3. H. -F. Chen , C. Wu , M. -C. Kuo , M. E. Thompson , and K. -T. Wong , J. Mater. Chem. , 22 , 9556 , 2012 . 4. P. G. Bomben , K. C. D. Robson , B. D. Koivisto , and C. P. Berlinguette , Coord. Chem. Rev. , 256 , 1438 , 2012 . 5. G. Jaouen and N. Metzler-Nolte (eds.), Medical Organometallic Chemistry , Springer , New York , 2010 . 6. G. M. Whitesides and J. Deutch , Nature , 469 , 21 , 2011 . 7. S. A. Marcott , J. D. Shakun , P. U. Clark , and A. C. Mix , Science , 339 , 1198 , 2013 . 8. R. H. Crabtree , Organometallics , 30 , 17 , 2011 . 9. M. L. Helm , M. P. Stewart , R. M. Bullock , M. Rakowski DuBois , D. L. DuBois , Science , 333 , 863 , 2011 ; J. D. Blakemore , N. D. Schley , G. Olack , C. D. Incarvito , G. W. Brudvig , and R. H. Crabtree , Chem. Sci., 2 , 94 , 2011 . 10. P. J. Chirik and K. Wieghardt , Science , 327 , 794 , 2010 . 11. K. -H. Ernst , F. R. W. P. Wild , O. Blacque , and H. Berke , Angew Chem-Int. Ed. , 50 , 10780, 2011 ; E. C. Constable and C. E. Housecroft , Chem. Soc. Rev. , 42 , 1429 , 2013 . 12. N. Selander and K. J. Szabo , Chem. Rev., 111 , 2048 , 2011 . 13. W. A. Nugent , Angew. Chem. Int. Ed. , 51 , 2 , 2012 . 14. J. Goodman , V. V. Grushin , R. B. Larichev , S. A. Macgregor , W. J. Marshall , and D. C. Roe , J. Am. Chem. Soc. , 131 , 4236 , 2009 . 15. L. C. Liang , Coord. Chem. Rev., 250 , 1152 , 2006 . 16. H. Bolvin Eur. J. Inorg. Chem., 2010 , 2221 . 17. F. Weinhold and C. R. Landis , Valency and Bonding , Cambridge University Press , New York 2005 ; G. Frenking and N. Fröhlich , Chem. Rev., 100 , 717 , 2000 . 18. B. E. Bursten , J. C. Green , N. Kaltsoyannis , M. A. MacDonald , K. H. Sze , and J. S. Tse , Inorg. Chem. , 33 , 5086 , 1994 . 19. V. Jonas , C. Boehme , and G. Frenking , Inorg. Chem. , 35 , 2097 , 1996 . 20. X. Solans-Monfort , C. Coperet , and O. Eisenstein , Organometallics , 31 , 6812 , 2012 . 21. G. J. Kubas , Metal Dihydrogen and σ -Bond Complexes , Kluwer/Plenum , New York , 2001 . 22. A. J. Cowan and M. W. George , Coord. Chem. Rev. , 252 , 2504 , 2008 . 23. T. Brewster , W. Ding , N. D. Schley , N. Hazari , V. Batista , and R. H. Crabtree , Inorg. Chem. , 50 , 11938 , 2011 . 38 INTRODUCTION

24. H. Taube , Pure Appl. Chem., 63 , 651 , 1991 . 25. N. M. West , J. A. Labinger , and J. E. Bercaw , Organometallics , 30 , 2690, 2011 . 26. J. R. Khusnutdinova , N. P. Rath , and L. M. Mirica , J. Am. Chem. Soc. , 132 , 7303 , 2010 . 27. U. Hintermair , S. W. Sheehan , A. R. Parent , D. H. Ess , D. T. Richens , P. H. Vaccaro , G. W. Brudvig , and R. H. Crabtree , J. Am. Chem. Soc. , 135 , 10837 , 2013 . 28. J. D. Blakemore , N. D. Schley , G. Olack , C. D. Incarvito , G. W. Brudvig , and R. H. Crabtree , Chem. Sci, 2 , 94 , 2011 . 29. R. H. Crabtree , New J. Chem. , 35 , 18 , 2011 .

PROBLEMS

1.1. How many isomers would you expect for a complex with an empirical formula corresponding to Pt(NH3 )2 Cl2 ? 1.2. What d n confi gurations should be assigned to the following and what magnetic properties—dia- or paramagnetic—are to be expected from the hexaqua complexes of Zn(II), Cu(II), Cr(II), Cr(III), Mn(II), and Co(II).

1.3. Why is R2 PCH2 CH2 PR2 so much better as a chelating ligand than 3 + R 2 PCH2 PR2 ? Why is H2 O a lower-fi eld ligand for Co than NH3 ? 1.4. How would you design a synthesis of the complex trans - [PtCl 2 (NH3 )(tu)], (the trans descriptor refers to the fact a pair of identical ligands, Cl in this case, is mutually trans), given that the trans effect order is tu > Cl > NH3 (tu = (H2 N)2 CS, a ligand that binds via S)?

1.5. Consider the two complexes MeTiCl3 and (CO)5 W(thf). Predict the order of their reactivity in each case toward the following sets of ligands: NMe 3 , PMe3 , and CO. 1.6. How could you distinguish between a square planar and a tetra- hedral structure in a nickel(II) complex of which you have a pure sample, without using crystallography?

1.7. You have a set of different ligands of the PR3 type and a large supply of (CO) 5 W(thf) with which to make a series of complexes (CO)5 W(PR3 ). How could you estimate the relative ordering of the electron-donor power of the different PR3 ligands? 1.8. The stability of metal carbonyl complexes falls off markedly as we go to the right of group 10 in the periodic table. For example, PROBLEMS 39

Cu complexes only bind CO weakly. Why is this? What oxidation state, of the ones commonly available to copper, would you expect to bind CO most strongly? 1.9. Low-oxidation-state complexes are often air sensitive (i.e., they react with the oxygen in the air) but are rarely water sensitive. Why do you think this is so? 5 1.10. MnCp 2 is high spin, while Mn(Cp*)2 (Cp* = η -C5 Me5 ) is low spin. How many unpaired electrons does the metal have in each case, and which ligand has the stronger ligand fi eld? 1.11. Why does ligand 1.18 bind as a clamshell with the Me and Cl sites mutually cis, and not in a coplanar arrangement with Me and Cl trans? 1.12. Make up a problem on the subject matter of this chapter and provide an answer. This is a good thing for you to do for subse- quent chapters as well. It gives you an idea of topics and issues on which to base questions and will therefore guide you in study- ing for tests.